Søren Andrea Just, MD, PhD, an associate professor at the University of Southern Denmark, presented groundbreaking findings at the recent American College of Rheumatology (ACR) Convergence 2024 conference, held from November 14-19 in Washington, DC. His research unveils that the performance of artificial intelligence (AI) in joint assessment can rival that of specialized rheumatologists, showcasing remarkable accuracy in evaluating joint disease activity.1,2
“We are witnessing a significant decline in the number of practicing rheumatologists in the US and Europe, while the patient population continues to grow,” Just commented during the conference, as reported by HCPLive®. “This presents a pressing challenge, which is why we explored innovative solutions like AI technology. If AI can deliver high-quality assessments of disease activity, particularly in areas where rheumatologists are scarce or non-existent, it could substantially improve patient care without necessitating an increase in healthcare personnel.”
One key study evaluated ARTHUR, a novel CE-marked fully automated ultrasound scanning system designed to capture ultrasound images of 22 hand joints. The AI system, DIANA, subsequently analyzes these images to assess and grade synovial hypertrophy (SH) and Doppler activity, following the globally recognized OMERACT-EULAR synovitis score (GLOESS).1
In terms of performance metrics, Just and his team discovered that ARTHUR and DIANA achieved a Kappa value of 0.39 [95% CI: 0.32–0.45] for SH, compared to 0.46 [95% CI: 0.41–0.52] when assessed by a rheumatologist. For Doppler activity, AI achieved a Kappa of 0.48 [95% CI: 0.41–0.55], in comparison to the rheumatologist’s Kappa of 0.45 [95% CI: 0.38–0.52].1
Combining the assessments across all joints in a binary framework (healthy vs. diseased), AI demonstrated an impressive agreement rate of 86.67% [95% CI: 69.28–96.24%] for SH and 83.33% [95% CI: 65.28–96.36%] for Doppler activity. In stark contrast, the rheumatologist’s assessment yielded a mere agreement rate of 53.33% [95% CI: 34.33–71.66%] for SH and 66.67% [95% CI: 47.19–82.71%] for Doppler activity.1
The second study investigated an AI model’s ability to evaluate greyscale and Doppler synovitis severity as well as osteophyte severity in hand joints, comparing its performance against that of human expert raters. When juxtaposed with a consensus score, the AI achieved a Kappa of 0.39 (95% CI: 0.35–0.44), a PEA of 51.77% (95% CI: 48.83–54.70%), a PCA of 91.03% (95% CI: 89.21–92.63%), a sensitivity rate of 46.19% (95% CI: 39.13–53.32%), and a specificity of 90.43% (95% CI: 88.35–92.25%) for SH.2
For assessing Doppler Activity, the AI demonstrated a Kappa of 0.61 (95% CI: 0.54–0.67), a PEA of 80.49% (95% CI: 77.51–83.22%), a PCA of 97.13% (95% CI: 95.69–98.18%), sensitivity of 67.31% (95% CI: 51.86–80.24%), and specificity of 96.29% (95% CI: 94.65–97.52%). In terms of grading osteophytes, the AI achieved a Kappa of 0.55 (95% CI: 0.46–0.63), with a PEA of 70.69% (95% CI: 65.57–75.45%), a PCA of 96.28% (95% CI: 93.70–98.01%), sensitivity of 56.43% (95% CI: 31.56–73.36%), and specificity of 95.36% (95% CI: 92.44–97.36%).2
“I truly hope these automated systems can facilitate rapid assessments that provide reassurance to patients, alleviating some of the pressures faced by rheumatologists while ensuring high-quality evaluations for those in need,” Just emphasized, underlining the potential of AI in transforming the landscape of rheumatologic care.
How is the ARTHUR ultrasound system integrated with AI to improve the evaluation of joint disease activity?
**Interview with Dr. Søren Andrea Just on AI in Rheumatology**
**Editor:** Thank you for joining us today, Dr. Just. Your recent presentation at the ACR Convergence 2024 conference highlighted some impressive findings regarding the use of artificial intelligence in joint assessment. Can you tell us about the significance of your research?
**Dr. Just:** Thank you for having me. I’m thrilled to share our findings. We demonstrated that AI can perform joint assessments with an accuracy that rivals that of specialized rheumatologists. This is crucial given the growing patient population and the declining number of practicing rheumatologists in the US and Europe. Using AI, we could potentially address the gap in healthcare access, especially in underserved areas.
**Editor:** That’s fascinating! One of the key technologies you discussed is the ARTHUR ultrasound scanning system. How does it work in conjunction with AI to evaluate joint disease activity?
**Dr. Just:** ARTHUR is a CE-marked, fully automated ultrasound system designed to capture images of 22 hand joints. Once these images are obtained, our AI system, DIANA, analyzes them to assess and grade synovial hypertrophy and Doppler activity based on the OMERACT-EULAR synovitis score. This allows us to standardize evaluations and reduce variability associated with manual assessments.
**Editor:** Your performance metrics are quite promising. Can you elaborate on the comparison between AI and rheumatologists in your study?
**Dr. Just:** Certainly. For synovial hypertrophy, ARTHUR and DIANA achieved a Kappa value of 0.39, while the rheumatologists had a Kappa of 0.46. For assessing Doppler activity, AI scored a Kappa of 0.48, compared to 0.45 for the rheumatologists. These results suggest that, while AI’s performance is very close, it is particularly effective in certain areas, indicating a complementary role rather than a replacement for human expertise.
**Editor:** It sounds like AI could play a vital role in the future of rheumatology. What are your thoughts on the challenges and the acceptance of AI in clinical practice?
**Dr. Just:** There are certainly challenges, including the need for robust integration of AI tools into existing workflows, addressing regulatory concerns, and ensuring that healthcare professionals are trained to use these technologies effectively. It’s important that we approach the integration of AI thoughtfully and ethically to enhance patient care while maintaining the human touch in medicine.
**Editor:** Thank you, Dr. Just, for sharing these insights. Your research certainly points to a future where AI could significantly enhance patient outcomes in rheumatology.
**Dr. Just:** Thank you for the opportunity to discuss our work. I’m optimistic about the future and excited about the potential of AI to improve healthcare delivery.
